The Reflex: MINT-AI

An AI-assisted Motivational Interviewing training tool that builds practitioner reflexes through simulated and role-play practice enhancing person-centered carewithout replacing human relationship.

Comment

Inspiration

The Reflex: MINT-AI was inspired by real-world experience in recovery and mental health peer support, where Motivational Interviewing (MI) is widely used but rarely practiced in a structured, reflective way. Many helpers are expected to “just know” how to respond well under pressure, even though decision-making is hardest in moments of stress, trauma, or emotional activation.

I wanted to explore whether AI could be used not as an authority or replacement for human judgment, but as a reflective training companion—one that helps people slow down, examine their responses, and learn how different choices shape outcomes. This project is also part of a broader concept called Anonymous Haven, which explores transparent, ethical AI systems that support trust, learning, and human dignity without surveillance or exploitation.

At its core, this project is about decision-making, economics, and trust: how people learn skills that reduce volatility, improve outcomes, and create long-term value for individuals and communities.

What it does

The Reflex: MINT-AI is an AI-assisted training tool for practicing Motivational Interviewing through reflection and feedback. Users can enter example responses or short dialogue snippets, and the system analyzes them using MI principles such as empathy, open-ended questions, reflections, and support for autonomy.

Instead of scoring or judging users, the tool provides qualitative feedback that helps users understand:

  • how their language may affect trust and engagement
  • how different responses can support or hinder behavior change
  • how reflective communication improves decision-making over time

The system is designed to be educational and non-punitive. It does not store personal data, make predictions about individuals, or enforce compliance. Its purpose is to help users build skill and awareness through practice.

How we built it

The project uses Google Gemini 3 as the core AI model for language analysis and feedback generation. A provider-based architecture allows switching between a mock provider (for reliable demos) and the real Gemini API.

The backend is built with Node.js and TypeScript, while the frontend demo is a lightweight HTML/JavaScript interface that allows users to test the system interactively. This structure keeps the system transparent and easy to audit, while demonstrating real-world AI integration.

Special care was taken to design the prompts and responses so the AI behaves as a reflective assistant rather than an authority figure. The goal was to demonstrate responsible AI use in a sensitive domain.

Challenges we ran into

One of the main challenges was balancing usefulness with ethics. It is easy for AI systems to become evaluative, prescriptive, or overly confident—especially in healthcare-adjacent settings. Designing prompts that encourage reflection without judgment required careful iteration.

Another challenge was ensuring the demo would be reliable and reproducible for judges. This was addressed by implementing a mock provider and clear setup instructions so the project can be run locally without API limitations.

Accomplishments that we're proud of

  • Successfully integrated Gemini 3 into a working, testable application
  • Built a provider architecture that supports both mock and real AI usage
  • Created a non-exploitative AI training tool focused on learning, not surveillance
  • Delivered a complete, professional demo with clear documentation
  • Aligned technical execution with ethical and economic design principles

What we learned

This project reinforced that AI can support better decision-making when it is designed with humility and transparency. Small improvements in communication skills can have large downstream effects—reducing conflict, increasing trust, and improving outcomes across systems.

We also learned that responsible AI design is not just about model performance, but about incentives, boundaries, and respect for human agency. Tools like The Reflex: MINT-AI can help people practice skills that generate long-term economic and social value, without turning human behavior into something to be controlled or exploited

What's next for The Reflex: MINT-AI

Future work includes expanding The Reflex: MINT-AI into guided simulations, scenario-based training, and additional reflective tools for peer support, healthcare, and education. As part of the broader Anonymous Haven concept, future iterations will explore transparent incentive design, privacy-preserving analytics, and ethical AI systems that support learning without surveillance or control.

Built With

Share this project:

Updates